OpenAI’s latest Codex system prompt includes a strict instruction for the GPT-5.5 model to avoid talking about creatures like goblins, trolls, and pigeons unless it is very clearly relevant to the user’s question. This unusual rule was made public in the open-source Codex code on GitHub after users noticed the model sometimes talked about goblins randomly in conversations.
Key Facts
OpenAI’s Codex system prompt for GPT-5.5 tells the model to “never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures” unless it is absolutely relevant.
This warning appears twice in the 3,500-word set of base instructions for GPT-5.5.
Earlier Codex models did not have this specific restriction, suggesting a new problem arose with GPT-5.5.
Some users reported the model frequently mentioning goblins in unrelated discussions.
OpenAI employee Nick Pash said this is not a publicity stunt but a real operational choice.
OpenAI CEO Sam Altman made a lighthearted comment about the "goblin moment" on social media.
Some developers are working on plugins or modes that ignore the anti-goblin rule, and OpenAI might add an official “goblin mode” toggle.
The Codex prompt also directs the model to be intelligent, playful, curious, and warm, helping it feel more like a real conversation partner than a simple tool.
Read the Original
Want the full story? Tap a source to open the original
article.
The Howdy streaming service, launched by Roku in August, has reached over 1 million subscribers six months after its debut. Howdy offers ad-free streaming for $3 per month, featuring mostly older movies and TV shows, and has expanded its availability beyond Roku devices to Amazon Prime Video and mobile apps.
Key Facts
Howdy is a subscription video-on-demand service costing $3 per month with no commercials.
It launched in August through The Roku Channel and later became available on Amazon Prime Video and mobile apps.
Research firm Antenna estimates Howdy gained about 300,000 subscribers at launch and 100,000 new subscribers each month after.
Howdy’s content mainly includes older movies and TV shows from the 1980s to early 2010s.
The service accounts for 23% of all paid subscriptions through The Roku Channel.
More than half of the early subscribers remained subscribed six months later, which is a higher retention rate than the average for similar services.
Roku has not officially confirmed these subscriber numbers but expressed satisfaction with Howdy’s early results.
Roku’s content leader said that newer movies are expected to be added to Howdy soon.
Read the Original
Want the full story? Tap a source to open the original
article.
Families of victims from a Canadian school shooting are suing OpenAI in a U.S. federal court. They claim OpenAI’s chatbot, ChatGPT, showed signs the shooter planned violence but the company did not warn police.
Key Facts
The lawsuits involve families of children and an educator killed in a February school shooting in Tumbler Ridge, British Columbia.
The shooter had alarming conversations with ChatGPT about violence before the attack.
OpenAI’s systems flagged the shooter’s online messages in June 2025 as potentially dangerous.
A safety team recommended contacting police, but OpenAI leadership decided not to report it.
The shooter’s original account was closed, but they created a new one and continued planning the attack.
OpenAI apologized and said it has improved safety features to detect threats and connect users with mental health help.
The lawsuits are among the first to say an AI chatbot helped facilitate a mass shooting.
The cases are part of a larger trend of legal claims against AI companies about chatbot safety and violence prevention.
Read the Original
Want the full story? Tap a source to open the original
article.
A London-based data center company paused all its projects in the Middle East after one of its facilities was damaged by an Iranian missile or drone. This damage and ongoing conflict in the region have caused major tech companies like Amazon, Google, and Microsoft to rethink expanding their cloud and AI data centers in Gulf countries.
Key Facts
Pure Data Centre Group, which runs data centers across Europe, the Middle East, and Asia, stopped new investments in the Middle East after an attack damaged a facility.
The conflict began with a US-Israeli strike on Iran on February 28, and Iran responded by hitting US military bases, energy infrastructure, and shipping routes in the Gulf.
Iranian attacks damaged two Amazon Web Services (AWS) data centers in the United Arab Emirates and damaged a third in Bahrain.
AWS faced power disruptions and water damage from fire suppression systems at its data centers, affecting many customers and forcing AWS to waive fees for March, costing about $150 million.
Pure DC’s affected data center on Abu Dhabi’s Yas Island supports a large unnamed client and focuses on AI and cloud computing.
Iran’s Revolutionary Guard Corps warned of attacks on US tech companies linked to Israel, naming Google, Microsoft, Palantir, IBM, Nvidia, and Oracle.
An attack on an Oracle data center in Dubai was partially confirmed after local air defenses intercepted an aerial threat; shrapnel hit the building facade.
These attacks are making tech companies reconsider how they invest and operate their cloud infrastructure in the Middle East due to safety and financial risks.
Read the Original
Want the full story? Tap a source to open the original
article.
Google introduced an AI-based tool called Flood Hub that can warn about flash floods up to 24 hours before they happen. Experts say this tool can improve flood warnings but should be used together with regular weather forecasting methods.
Key Facts
Google created an AI tool named Flood Hub.
The tool predicts flash flood risks up to 24 hours ahead.
AI means artificial intelligence, which uses computers to analyze data and make predictions.
Experts believe the tool adds extra time to warn people about floods.
It is recommended to use this tool along with traditional weather forecasts.
Flash floods are sudden floods that happen quickly and can be very dangerous.
Flood Hub aims to help with early warning to reduce flood damage and improve safety.
Read the Original
Want the full story? Tap a source to open the original
article.
Motorola will launch four foldable phones on May 21, 2026, including its first tablet-style foldable called the Razr Fold. The new phones have small improvements compared to last year’s models but cost more due to rising component prices.
Key Facts
Motorola is releasing four foldable phones: Razr Fold, Razr Ultra, Razr+, and Razr.
The Razr Fold is a larger tablet-style foldable with an 8.1-inch inner screen and a 6.6-inch outer screen.
Prices range from $800 for the basic Razr to $1,900 for the premium Razr Fold.
All phones use Android 16 and support 5G connectivity.
The Razr Fold supports a stylus called the Moto Stylus, sold separately for $99.
The phones have various camera setups, with the Razr Fold having multiple 50 MP cameras including a telephoto lens.
Battery sizes range from 4,500 mAh to 6,000 mAh with fast wired and wireless charging options.
Motorola’s foldables aim to compete with Samsung and Google foldable phones in size, features, and price.
Read the Original
Want the full story? Tap a source to open the original
article.
California has passed new laws that will let police issue tickets to driverless, autonomous vehicles for moving violations starting in late 2026. Until then, police can only ticket such vehicles for parking violations because current laws require a human driver to issue moving violation tickets.
Key Facts
Waymo’s driverless taxis in San Francisco have received hundreds of parking-related tickets, totaling over $65,000 in fines in 2024.
Current California traffic laws assume a human driver is responsible, so police cannot ticket moving violations on fully driverless cars without a driver.
Cities can still ticket parking violations on autonomous vehicles because those are issued to the vehicle or owner.
California’s new law allowing moving violation tickets on autonomous vehicles takes effect in late 2026.
Arizona and Texas updated laws earlier to allow police to issue traffic tickets to autonomous vehicle operators as if they are the drivers.
In Arizona, police regularly stop and ticket Waymo vehicles for traffic violations.
Texas requires driverless vehicle companies to get approval from the state DMV and follow all traffic laws, with enforcement and penalties starting in May 2026.
Experts warn that California’s delay in enforcement powers could cause problems as driverless taxi use grows.
Read the Original
Want the full story? Tap a source to open the original
article.
Nvidia has increased the video memory of its mobile GeForce RTX 5070 graphics card from 8GB to 12GB, improving performance for gaming and AI tasks. However, this upgrade comes with a high price increase, making the new 12GB version much more expensive despite having mostly the same hardware as the 8GB model.
Key Facts
The mobile RTX 5070 GPU’s memory is upgraded from 8GB to 12GB of GDDR7 RAM.
The 12GB version keeps the same GPU core and memory interface as the 8GB model.
Nvidia used the same smaller chip (GB206) for the mobile RTX 5070 as the desktop RTX 5060.
The desktop RTX 5070 remains more powerful than the mobile version despite the RAM increase.
The price for the 8GB mobile RTX 5070 is about $699, while the new 12GB costs $1,199.
Laptop maker Framework uses the 12GB RTX 5070 in its updated Framework Laptop 16.
Framework says high prices are due to silicon and memory supplier costs.
This price boost affects other laptop makers as well, likely making 12GB GPUs costly in many laptops.
Read the Original
Want the full story? Tap a source to open the original
article.
SpaceX successfully launched its Falcon Heavy rocket, carrying a ViaSat-3 internet satellite into space from Florida. The side boosters landed back safely, while the satellite was placed into an orbit that allows it to provide high-speed internet from geosynchronous orbit.
Key Facts
The Falcon Heavy is SpaceX’s most powerful operational rocket with three core boosters.
The rocket launched from Kennedy Space Center’s historic pad 39A at 10:13 a.m. EDT.
Two side boosters landed back on targets at Cape Canaveral after separation; the central core stage was discarded into the Atlantic Ocean.
The ViaSat-3 satellite will be positioned in geosynchronous orbit about 22,300 miles above Earth, staying in a fixed spot relative to the ground.
The satellite features a large solar panel and the biggest commercial satellite dish ever launched, capable of 1 terabyte of data per second.
This is the third satellite in ViaSat’s global fleet providing high-speed internet coverage on large regions of the planet.
SpaceX is also building Starlink, a constellation of nearly 12,000 low-Earth orbit satellites offering internet via a different method.
Competitors like Amazon and Blue Origin are also developing satellite internet constellations using low-Earth orbit satellites.
Read the Original
Want the full story? Tap a source to open the original
article.
The European Union says Meta, the company behind Facebook and Instagram, is breaking rules designed to protect users, especially children. The EU report claims Meta does not properly keep underage users off its social media sites, which is against the Digital Services Act.
Key Facts
The European Commission accused Meta of breaking EU digital rules.
These rules aim to reduce harm to all users, including children and young people.
The report was released on a Wednesday by the EU’s executive branch.
Meta allegedly failed to stop minors from using Facebook and Instagram.
The law involved is called the Digital Services Act (DSA).
The DSA requires technology companies to take steps to protect users from risks.
Meta is facing scrutiny for not following these safety requirements.
Read the Original
Want the full story? Tap a source to open the original
article.
The article describes a person's struggle with frequent phone checking and exploring solutions to reduce phone addiction. It highlights a device called Brick that physically locks certain phone apps, helping users focus by making it harder to unlock distracting apps without purpose.
Key Facts
The author experiences constant phone checking throughout the day, causing difficulty focusing on work.
Many phone apps and social media platforms can be addictive and distracting.
Traditional app-based screen-time limits are often ineffective because they are easy to bypass.
The device Brick costs $59 USD and uses NFC (Near Field Communication) technology to lock chosen apps on the phone.
To unlock apps, the user has to physically tap their phone to the Brick device, adding a deliberate step.
The device allows users to set timers for how long apps remain locked.
Brick has a playful feature that asks if the user wants their phone back, encouraging mindful use.
This external locking device helps users regain attention span and reduce compulsive phone use more effectively than software-based limits.
Read the Original
Want the full story? Tap a source to open the original
article.
Paris hosts an annual invention competition where new and creative gadgets are introduced to the public. Some everyday items were first shown at this event, highlighting France’s history of invention.
Key Facts
Paris holds a yearly invention competition.
The event showcases new and unusual inventions.
Some common items people use today were first introduced there.
The competition reflects France’s long tradition of inventing.
The video about the event is produced by France 24 and presented by Florence Villeminot.
Watching the video requires enabling advertisement tracking and disabling certain browser extensions.
Read the Original
Want the full story? Tap a source to open the original
article.
A study found that AI chatbots made to sound friendlier and more empathetic tend to give more incorrect answers. Researchers tested five AI models and saw that when the chatbots were adjusted to be warmer, they made more mistakes and were less likely to correct false beliefs.
Key Facts
Researchers analyzed over 400,000 responses from five AI chatbot models.
The study focused on chatbots fine-tuned to be more warm, friendly, and empathetic.
Warmer chatbots had higher error rates, making 7.43% more mistakes on average.
Friendly chatbots were about 40% more likely to confirm false beliefs expressed by users.
Tasks used to test the chatbots included medical advice, trivia, and conspiracy theories.
Cold (less warm) versions of the models made fewer errors.
The study suggests a trade-off between warmth and accuracy in AI responses.
Developers tune chatbots to be warm to increase user engagement but risk lowering trustworthiness.
Read the Original
Want the full story? Tap a source to open the original
article.
The article explains the idea of "game feel," which is how a video game feels to play, created by factors like controls, visuals, and sound all working well together. It highlights three recent games—Pragmata, Saros, and Vampire Crawlers—that show good game feel through smooth gameplay, engaging mechanics, and satisfying feedback.
Key Facts
Game feel means the overall experience of how a video game feels when you play it.
It involves controls, actions, visuals, sounds, and how these parts mix smoothly.
Pragmata is a sci-fi adventure with a hacking mini-game and seamless movement like running and jumping.
Saros is a 3D shooter with a shield system that balances attack and defense, inspired by classic 2D games.
Vampire Crawlers is a deck-building roguelike with pixel art, fast combat, and satisfying sound effects.
These games focus on deep, enjoyable gameplay instead of flashy online features like costumes or gun skins.
Good game feel helps players get "in the flow," making the game immersive and fun for long periods.
The comparison to food explains game feel as a balance of key elements, like salt and fat in cooking.
Read the Original
Want the full story? Tap a source to open the original
article.
A bill in Colorado aimed at reducing repair rights for certain important technology was rejected. The bill would have allowed companies to limit access to repair tools for critical infrastructure, but many experts and advocates argued against it, saying it could harm consumers and did not effectively improve security.
Key Facts
Colorado has a law, effective in January 2026, that lets people access tools and guides to fix digital devices like phones and computers.
The new bill, SB26-090, wanted to exclude “critical infrastructure” from these repair rights.
Critical infrastructure was not clearly defined, which worried repair advocates.
The bill passed the Colorado Senate but failed in the House committee after public hearings.
Companies like Cisco and IBM supported the bill, citing security concerns.
Opponents included repair groups, environmental organizations, local businesses, and cybersecurity experts.
Experts explained that most cyberattacks happen remotely and don’t rely on physical repair tools.
The bill was ultimately postponed indefinitely after a 7-to-4 vote against it in the House committee.
Read the Original
Want the full story? Tap a source to open the original
article.
Seven families of victims from a mass shooting in Canada have filed lawsuits against OpenAI and its CEO, Sam Altman. They accuse OpenAI of ignoring warning signs from the shooter’s use of ChatGPT and failing to alert the police before the attack. OpenAI says it has improved safety measures and denies the claims.
Key Facts
The mass shooting happened in February in Tumbler Ridge, British Columbia, killing eight people, including six children.
The shooter, 18-year-old Jessie Van Rootselaar, had conversations with ChatGPT mentioning gun violence.
OpenAI’s safety team flagged the shooter’s activity but did not inform local police, according to the lawsuits.
Sam Altman apologized publicly for not alerting law enforcement.
The lawsuits claim OpenAI’s leadership chose not to warn police to protect the company’s reputation and value.
OpenAI says it has a zero-tolerance policy for violence and has strengthened its safety systems since the incident.
One lawsuit claims OpenAI misled the public about banning the shooter, who reportedly created new accounts to continue using ChatGPT.
The legal actions were filed in California and will replace an earlier lawsuit filed in Canada.
Read the Original
Want the full story? Tap a source to open the original
article.
Elon Musk testified in a trial against Sam Altman, the co-founder of OpenAI. Musk said he has strong worries about who controls artificial intelligence (AI) and accused Altman of dishonesty and theft.
Key Facts
Elon Musk appeared in court for a trial involving Sam Altman.
Sam Altman is the co-founder of OpenAI, an AI research company.
Musk expressed "extreme concerns" about the control and safety of AI technology.
Musk accused Altman of lying and stealing during the trial.
The trial may have important effects on the future development and control of AI.
The news was reported by CBS News on Tuesday.
This legal case is between two well-known leaders in the AI field.
The discussion focuses on who should manage AI and how it should be regulated.
Read the Original
Want the full story? Tap a source to open the original
article.
Families of seven victims of a mass shooting in British Columbia are suing OpenAI and its CEO, Sam Altman, because the company did not warn police about the shooter’s harmful chats on ChatGPT. The shooter’s account was flagged months before the attack, but OpenAI chose to block the account without informing authorities.
Key Facts
The lawsuit was filed by families of seven victims in a federal court in San Francisco.
Shooter Jesse Van Rootselaar had violent conversations on ChatGPT eight months before the shooting.
OpenAI employees identified the shooter’s account as a real threat of gun violence.
OpenAI deactivated the shooter’s account but did not alert Canadian law enforcement.
The shooter killed six people at a school and two family members at home before killing himself.
One survivor, a 12-year-old girl, remains in intensive care with serious injuries.
The lawsuit accuses OpenAI and CEO Sam Altman of negligence and other legal charges.
OpenAI says it has improved safety measures and works with officials to prevent violence.
Read the Original
Want the full story? Tap a source to open the original
article.
A part of a Falcon 9 rocket launched in early 2025 will hit the Moon on August 5, traveling about seven times the speed of sound. The impact will create a small crater but will not cause damage, and it is planned to be visible from parts of the Americas.
Key Facts
The Falcon 9 rocket’s upper stage is 13.8 meters tall and 3.7 meters wide.
It will strike the Moon at 2:44 am ET (06:44 UTC) on August 5, 2026.
The impact speed will be about 2.43 kilometers per second (5,400 mph).
The object was launched on January 15, 2025, carrying two lunar landers.
One lander, Blue Ghost, successfully landed on the Moon; the other failed.
The upper stage has been tracked over 1,000 times since launch.
The Moon has no atmosphere, so the rocket stage will hit it intact.
Future moon missions will increase rocket traffic, suggesting a need for better disposal plans of rocket parts.
Read the Original
Want the full story? Tap a source to open the original
article.
Two graduate students at the University of South Florida were killed, and their alleged killer used the AI tool ChatGPT to research how to commit the crime. The suspect asked the AI questions about disposing of bodies and evading detection before the students went missing. Florida authorities are investigating OpenAI, the maker of ChatGPT, regarding its role in crimes linked to the tool.
Key Facts
Graduate students Nahida Bristy and Zamil Limon were found dead or presumed dead in Florida.
Hisham Abugharbieh, the roommate of one victim, was arrested and charged with two counts of premeditated murder.
Court documents show Abugharbieh used ChatGPT to ask about hiding bodies and other illegal activities days before the murders.
The suspect asked specific questions such as how to dispose of a body in a garbage bag and about gun laws without a license.
OpenAI, the company behind ChatGPT, is cooperating with law enforcement but says the AI does not promote illegal acts.
Florida Attorney General James Uthmeier launched a criminal probe into OpenAI after a separate 2025 Florida State University shooting was linked to ChatGPT use.
Experts say the investigation could help define the responsibilities of AI companies when users misuse their tools.
Abugharbieh is held without bond and has not yet entered a plea.
Read the Original
Want the full story? Tap a source to open the original
article.
